Gaussian Mixture Model for image classification using wavelet coefficients

I am trying to generate the Gaussian Mixture Model for a subband from the distribution of wavelet coefficients. But i think there is some mistake in my code regarding initialization of the Gaussian Mixture Model

X = mat_cell;num_dim=2;num_clus = 2; % number of mixtures/clusters

        [counts,binLocations] = imhist(X);
        stem(binLocations, counts, 'MarkerSize', 1 );
        xlim([-1 1]); 
        % inital kmeans step used to initialize EM
        rng('default');
        data=reshape(X,[],num_dim);
        [kmeanscid,cInd.mu] = kmeans(data, num_clus,'MaxIter', 75536);
        data_k_1=data(kmeanscid==1);
        data_k_2=data(kmeanscid==2);
        a = diag(cov(data_k_1'));
        b=diag(cov(data_k_2')); 
        if a==0
            a=0.1;
        elseif b==0
            b=0.1;   
        end
        initialsigma = cat(3,[a,a],[b,b]);  
        %disp(initialsigma);
        cInd.Sigma=initialsigma;
        % fit a GMM model
        options = statset('MaxIter', 75536); 
        gmm = fitgmdist(data, num_clus,'Start',cInd,'CovarianceType','diagonal','Regularize',1e-5,'Options',options);